LCF (lexical friction coefficient) - optimization - Kinda Long, but great!
Despite my vast (nay, encyclopedic) knowledge
and understanding of all aspects of the Internet, one thing puzzles me. That is
how I can send a message to a group and have it appear literally within
seconds, and then send another which will take more than 24 hours to appear.
Puzzled Los Gatos Sciolist
Ah... I'm happy to report that you have come
to the right place for the answer to this deep question.
Before I reveal the cause of the common
phenomenon you're wondering about, though, I'd like to point out some other
quirky behaviors that you may have noticed.
.. Some days your car starts on the first turn
of the starter. Some days it doesn't start at all.
.. Some recent nights have been brilliantly
lighted by the full moon. Tonight I've waited and waited, but all I got was
These have nothing to do with why one message
is transmitted immediately while another takes 24 hours.
The reason is complex, and we technologists
don't often expect even to hear such sophisticated questions from those outside
the inner circle (HAHAHAHAHA WE KNOW EVERYTHING), and many of us are loathe to
reveal the hidden cause.
But you seem trustworthy, so...
Look at your keyboard. Notice how the keys are
all out of order? You'd think they'd be in alphabetical order, wouldn't you?
But no, they are arranged in an odd pattern called QWERTY, originally devised
by a typewriter manufaturer to slow typists down to the point where his
machines wouldn't jam. Imagine, now, when you send a message down the wires,
how differently the many routers and interfaces that the message goes through
are affected by different juxtapositions of letters in your message. Just as a
modest change in the original position of a chess problem has a dramatic effect
on the time required to solve it, the tiniest change in the arrangement of
letters in your message - often not even noticeable to any but the expert eye,
and even then only with complex measurement equipment - can wreak havoc with
every interface the message must pass through. Imagine you had a car wider than
the normal freeway lane. Going through interchanges would be a particular
trial; how quickly you could pass through would depend on the amount of other
traffic, the number of odd-shaped oncoming cars, and many other factors -- much
too complex to summarize quickly. But I'm sure you get the idea. And just as if
you drove through many interchanges in your odd-shaped car you could be delayed
dramatically, changes in the letter composition of your message slow it down
every time it goes through a router, the internet's interchange.
The letters W and M are particularly noxious
in this way. If they happen to fall within the same word, as in women, or if
multiples of them fall within a word, as in mammal, or wow, their retardant
effect is in fact squared; this was first proved by Von Neumann in 1944;
although certain notes of Ada Lovelace in 1861 indicate that she, too, had the
The vowels, on the other hand, particularly I
and O, are quite slippery and can speed up the trip of your message through a
router; in fact, an I almost cancels an M, and words with many I's and O's,
such as oil, lion, noise, and onion, can have a remarkable accelerating effect.
These are only the extreme cases. Each letter,
and in fact each key, has its own lexical friction coefficient (LFC), which
often depends on the relationship of the letter to other letters in the word
and to other words in the message. LFC tables were originally compiled by
Hollerith in 1901, for use in his famous Census-tabulating work, but were not
made available to the general public until IBM brought out the 407 tabulating
machine in the mid-thirties, and published a full set of lexical friction data
in the documentation that was issued with the machine. Later, in 1962, when IBM
first produced the selectric typewriter, new LFC tables had to be constructed;
these were made available in technical libraries.
Depending on the net lexical friction of a
message, the transit time of a message through a router can differ by as much
as a factor of fifty. This in itself is hardly sufficient to explain the
difference between instant delivery and 24-hour delivery, however. The biggest
part of the effect is a second-order result of high-LFC messages passing
through routers. Just as when a stream slows down it deposits much more silt
and other sediment on its bed, a high-lfc message, traveling slowly through a
router, leaves what amounts to arterial plaque in the routers optical fiber
connections. Optical fiber builds up LFC-related plaque anyway, but normally so
slowly that fibers don't have to be cleaned or changed for years. However, a
chance confluence of many high-LFC message can deposit so much LFC plaque in
the fiber connections of a router that the router can be totally disabled. Even
if the router is not put completely out of service by fiber plaque, it can
transmit mesages so slowly that many recipient protocol managers conclude
correctly that their correspondents have failed, and request retransmission.
Thus high-LFC messages not only move more slowly through the internet, but
actually raise the internet's traffic load while they do so.
This issue has been studied in great detail by
my erstwhile employer, whose interest in fiber plaque, LFC aggregation, and the
resulting internet congestion is so high that it has formed a special task
force to study the matter and recommend solutions within a year. I fully
expect, however, that since the matter is dependent on keyboard design
originally, these studies will probably result in little improvement, and once
again we will be left anxiously awaiting the next-faster generation of optics,
routers, and computers, meanwhile helplessly floundering in a stew of such
technical complexity that only the few can comprehend it.
I would suggest that to improve your
transmission times you should begin by tabulating the letter counts in your
messages, and correlate them with message delivery delays. This technique is
crude, but should give you a rough idea of what to expect. If your needs go
beyond manual counting, you can find any number of lexical friction coefficient
analysis programs in the commercial world, replete with graphic interfaces and
I'm glad to have been of service in this
matter, and will make myself available for further questions as they occur to
Received from Teresa's Jokers.
The Good, Clean Funnies List: Good, clean funnies five times a week, FOR FREE!
Added November 13, 1999
-=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=-
Back to Top
Back to Computer and Internet Humor